diff --git a/imperative/python/megengine/amp/autocast.py b/imperative/python/megengine/amp/autocast.py
index 03163034ded97199848cefc4e2292f9069101ccc..eeb23620e143bc8087094139e5cba707a1306eab 100644
--- a/imperative/python/megengine/amp/autocast.py
+++ b/imperative/python/megengine/amp/autocast.py
@@ -15,14 +15,14 @@ class autocast:
     A class to control autocast mode for amp as a context manager or a decorator.
 
     :param enabled: Whether autocast mode is enabled.
-    :low_prec_dtype: Set amp autocast mode's lower precision dtype. It will change the
-    target dtype in tensor casting for better speed and memory. Default: float16.
-    :high_prec_dtype: Set amp autocast mode's higher precision dtype. It will change the
-    target dtype in tensor casting for better precision. Default: float32.
+    :param low_prec_dtype: Set amp autocast mode's lower precision dtype. It will change
+        the target dtype in tensor casting for better speed and memory. Default: float16.
+    :param high_prec_dtype: Set amp autocast mode's higher precision dtype. It will
+        change the target dtype in tensor casting for better precision. Default: float32.
 
     Examples:
 
-    ..code-block::
+    .. code-block::
 
         # used as decorator
         @autocast()
diff --git a/imperative/python/megengine/core/tensor/amp.py b/imperative/python/megengine/core/tensor/amp.py
index 4f8e49a36e39c3c399a0e7079b81c25527e28ecf..9d54927b59fe049f099f0bc173b1eba08ff56ac5 100644
--- a/imperative/python/megengine/core/tensor/amp.py
+++ b/imperative/python/megengine/core/tensor/amp.py
@@ -17,7 +17,7 @@ def enabled(mod):
 
     Examples:
 
-    ..code-block::
+    .. code-block::
 
         import megengine as mge
         mge.amp.enabled = True
@@ -40,7 +40,7 @@ def high_prec_dtype(mod):
 
     Examples:
 
-    ..code-block::
+    .. code-block::
 
         import megengine as mge
         mge.amp.high_prec_dtype = "float32"
@@ -63,7 +63,7 @@ def low_prec_dtype(mod):
 
     Examples:
 
-    ..code-block::
+    .. code-block::
 
         import megengine as mge
         mge.amp.low_prec_dtype = "float16"
diff --git a/imperative/python/megengine/quantization/qconfig.py b/imperative/python/megengine/quantization/qconfig.py
index 5050045c2f62f852fe48936e02c21aba28b09f4a..6d5c476a992b66807f2ba61084e4f2bde1207d60 100644
--- a/imperative/python/megengine/quantization/qconfig.py
+++ b/imperative/python/megengine/quantization/qconfig.py
@@ -28,7 +28,7 @@ class QConfig(
     )
 ):
     r"""
-    A config class indicating how to do quantize toward :class:`~.QATModule`'s
+    A config class indicating how to do quantize toward :class:`~.QATModule` 's
     ``activation`` and ``weight``. See :meth:`~.QATModule.set_qconfig` for detail usage.
 
     :param weight_observer: interface to instantiate an :class:`~.Observer` indicating
diff --git a/imperative/python/megengine/tensor.py b/imperative/python/megengine/tensor.py
index 8b09a1d180552d21468009f8bc443c2bc93c747b..2476355745bb39e3006cbbcd3c0b1bc448ab6419 100644
--- a/imperative/python/megengine/tensor.py
+++ b/imperative/python/megengine/tensor.py
@@ -30,6 +30,7 @@ class Tensor(_Tensor, ArrayMethodMixin):
     A tensor object represents a multidimensional, homogeneous array of fixed-size items.
 
     :param data: The value of returned Tensor.
+    :type data: Tensor, :class:`~.numpy.ndarray`, :class:`list` or python number.
     :param dtype: The dtype of returned Tensor. Uses data's dtype if not specified.
     :param device: The desired device of returned Tensor. Uses :func:`get_default_device` if not specified.
     :param is_const: Whether make it a ``ImutableTensor`` in tracing mode.
@@ -43,7 +44,7 @@ class Tensor(_Tensor, ArrayMethodMixin):
 
     def __new__(
         cls,
-        data: Union["Tensor", np.ndarray, list, "scalar"] = None,
+        data: Union["Tensor", np.ndarray, list, int, float] = None,
         dtype: np.dtype = None,
         device: str = None,
         is_const: bool = False,
@@ -76,7 +77,7 @@ class Tensor(_Tensor, ArrayMethodMixin):
 
     def __init__(
         self,
-        data: Union["Tensor", np.ndarray, list, "scalar"],
+        data: Union["Tensor", np.ndarray, list, int, float],
         dtype: np.dtype = None,
         device: str = None,
         is_const: bool = False,